Today Grizzly patch reviews are essentially a manual process involving V+2/C+2
Likewise, the deployment process is simple. Puppet updates a copy of the repo on the grafana server, and grr apply is run by the deployer to make the change live (in a nutshell puppet agent -t; grr apply).
Lots of room for improvements!
A few ideas to further explore:
- Initial jenkins scaffolding/integration
- Jsonnet lint/parse common entry points, e.g. slo_dashboards.jsonnet
- Automatically execute grr testing commands, these inherently perform parsing too potentially making the above moot. (note: grr needs api access, but perhaps allowing jenkins a read-only key against our public grafana instance is workable)
- grr snapshot (generates would-be dashboards a throwaway copies, useful for seeing exactly what will be deployed before deploying)
- grr diff (similar to PCC, shows what would be changed vs current production state)
- Gate/submit style deployment automation